gitaiflow 0.1.2

AI-assisted git diff intelligence and engineering change summary platform for Django and Python projects.

AIAssist : gitaiflow

Maintained by

https://djangoplay.org

AI-Assisted Git Intelligence & Engineering Analysis Platform

gitaiflow is a local-first AI-assisted engineering intelligence platform for:

• Git diff analysis
• engineering change summaries
• pull request summaries
• architectural analysis
• Django-aware change detection
• hierarchical diff summarization

The platform combines:

• deterministic Python analysis
• Git intelligence
• local LLM synthesis
• architecture-aware summarization

to generate high-quality engineering summaries from Git changes.

---

Philosophy

gitaiflow follows several core principles:

• local-first execution
• privacy-preserving workflows
• deterministic preprocessing
• AI-assisted synthesis
• architecture-aware engineering analysis
• no cloud dependency required

The system intentionally avoids:

• agent framework complexity
• vector databases
• autonomous orchestration
• unnecessary infrastructure

---

Features

Current Features

• recursive Git diff analysis
• single-file diff analysis
• file-level diff generation
• folder-level aggregate summaries
• compact GitHub-style summaries
• detailed engineering summaries
• aggregate-only summary generation
• Django-aware workflows
• local Ollama integration
• per-file markdown summaries
• aggregate engineering summaries
• GitHub/GitLab remote support
• configurable artifact output directory
• configurable git remote selection
• configurable base branch selection
• configurable model family selection
• configurable model size selection
• configurable path skipping
• local-first execution
• sensitive token filtering
• deterministic preprocessing
• prompt caching for faster execution
• interrupt-safe execution handling
• tracked and untracked file diff support

---

Prerequisites

1. Install Ollama

Official website:

https://ollama.com

Verify installation:

ollama --version

---

2. Pull Recommended Model

ollama pull llama3.2:3b

Alternative recommended models:

ollama pull qwen2.5-coder:7b

ollama pull mistral:7b

---

3. Start Ollama

ollama serve

---

Installation

Option 1 — Install from PyPI

pip install gitaiflow

Verify installation:

gitaiflow --help

---

Option 2 — Install from Source

git clone https://gitlab.com/codefleet-labs/aiassist.git

cd aiassist

pip install -e .

For Mac Users

pip install -e ".[dev]"

---

Quick Start

Analyze Entire Folder

gitaiflow \
    --path mailer/

Generates:

change-summary/
├── cache/
├── diff/
├── json/
├── markdown/
└── metadata/

---

Analyze Single File

gitaiflow \
    --path users/views/ui/logout.py

---

Generate Detailed Engineering Summary

gitaiflow \
    --path users/views/ui/logout.py \
    --full

---

Generate Aggregate Summary Only

gitaiflow \
    --path mailer/ \
    --aggregate

---

Re-Generate Summary From Existing Diff

gitaiflow \
    --target logout

---

Supported CLI Arguments

Argument	Description
--path	File or folder path to analyze
--target	Re-generate summary from existing diff
--aggregate	Generate aggregate summary only
--full	Generate detailed engineering summary
--remote	Override git remote name
--base-branch	Override git base branch
--model	Override model family
--size	Override model size
-o, --output-dir	Override artifact output root directory
--skip	Additional files/directories to skip

---

CLI Examples

Default Artifact Output

gitaiflow \
    --path mailer/

Generates:

change-summary/
├── cache/
├── diff/
├── json/
├── markdown/
└── metadata/

---

Custom Artifact Output Directory

gitaiflow \
    --path mailer/ \
    -o artifacts/

Generates:

artifacts/
├── cache/
├── diff/
├── json/
├── markdown/
└── metadata/

---

Use Specific Git Remote

gitaiflow \
    --path mailer/ \
    --remote github

---

Use Specific Base Branch

gitaiflow \
    --path mailer/ \
    --base-branch develop

---

Override Both Remote And Branch

gitaiflow \
    --path mailer/ \
    --remote upstream \
    --base-branch master

---

Use Specific Model

gitaiflow \
    --path . \
    --model qwen2.5-coder \
    --size 7b

---

Skip Additional Paths

gitaiflow \
    --path . \
    --skip migrations legacy tests

---

Skip Multiple Directories

gitaiflow \
    --path . \
    --skip \
        aiassist/config/ \
        tests/ \
        docs/

---

Generate Aggregate Summary Only

gitaiflow \
    --path . \
    --aggregate

---

Artifact Generation

By default, gitaiflow generates artifacts inside:

./change-summary/

relative to the current execution directory.

Generated structure:

change-summary/
└── 2026/
    └── 2026-05-16/
        └── v0.1.0/
            ├── cache/
            ├── diff/
            ├── json/
            ├── markdown/
            └── metadata/

Artifact directories are grouped by:

• generation year
• generation date
• {app_version}

This allows:

• release-day traceability
• historical summary preservation
• multiple generation runs per day
• isolated summaries across gitaiflow versions

Directories are automatically:

• created if missing
• reused if existing
• regenerated safely during execution

The system clears only generated runtime artifacts:

• diff/
• markdown/

while preserving:

• cache/
• json/
• metadata/

This prevents stale summaries while preserving reusable metadata.

Custom output root may be specified using:

-o artifacts/

---

Git Comparison Strategy

gitaiflow compares changes against:

<remote>/<base-branch>

Examples:

origin/main
github/main
upstream/develop
gitlab/main

Default values:

remote       = origin
base-branch  = main

Both may be overridden dynamically:

gitaiflow \
    --path mailer/ \
    --remote github \
    --base-branch develop

If remote or branch does not exist, execution halts safely with validation errors.

---

Security & Sensitive Data Handling

gitaiflow automatically redacts common sensitive credential patterns from generated diffs.

Current protected patterns include:

• SECRET_KEY
• API_KEY
• PRIVATE_KEY
• ACCESS_KEY
• CLIENT_SECRET
• secret_key
• api_key
• private_key
• access_key
• client_secret

Sensitive lines are replaced with:

[REDACTED SENSITIVE CONTENT]

The filtering system intentionally avoids false positives for:

• imports
• comments
• variable names like token_count
• helper functions

---

Generated Artifact Structure

change-summary/

├── cache/
├── diff/
├── json/
├── markdown/
└── metadata/

Generated artifacts include:

Directory	Purpose
diff/	Generated Git diffs
markdown/	AI-generated markdown summaries
cache/	Cached runtime artifacts
json/	Structured machine-readable outputs
metadata/	Internal execution metadata

---

Prompt System

gitaiflow ships with multiple prompt templates:

• compact summaries
• full engineering summaries
• aggregate summaries
• aggregate engineering summaries

Prompt templates are preloaded into memory during startup for:

• fail-fast validation
• reduced filesystem overhead
• stable long-running execution

---

Supported Summary Modes

Compact Summary

Designed for:

• PR descriptions
• quick reviews
• GitHub/GitLab summaries
• concise engineering updates

Example:

gitaiflow \
    --path . \
    --compact

---

Full Engineering Summary

Designed for:

• architecture reviews
• senior engineering analysis
• implementation audits
• detailed engineering reporting

Example:

gitaiflow \
    --path . \
    --full

---

Recommended Models

Model	Usage
llama3.2:3b	lightweight default
llama3.1:8b	higher-quality summaries
qwen2.5-coder:7b	strong code reasoning
mistral:7b	fast summarization
deepseek-coder:6.7b	code-aware analysis

---

Example Large Repository Analysis

gitaiflow \
    --path . \
    --target aiassist/ \
    --o summary/dir/ \
    --remote gitlab \
    --model qwen2.5-coder \
    --size 7b \
    --skip aiassist/config/

---

Development Setup

Clone repository:

git clone https://gitlab.com/codefleet-labs/aiassist.git

cd aiassist

Install development environment:

pip install -e .[dev]

Run linting:

ruff check .

Run tests:

pytest

---

PyPI Publishing

Build package:

python -m build

Upload to TestPyPI:

twine upload --repository testpypi dist/*

Upload to PyPI:

twine upload dist/*

---

Planned AI Provider Support

Future releases will support:

Provider	Status
Ollama	Current Default
OpenAI GPT-5	Planned
Claude	Planned
Gemini	Planned
DeepSeek API	Planned
OpenRouter	Planned
Azure OpenAI	Planned

Ollama will remain the default local-first provider.

---

Versioning

This project follows Semantic Versioning.

0.x → active development
1.x → stable production releases

Current Version

v.0.1.0

---

License

MIT License.

See LICENSE.

---

Planned Release Features

See TODO.

👤 Maintainer

DjangoPlay

[e-mail]

Actively maintained (updated May 16, 2026)

---

🌟 Contributing

Pull requests welcome. Follow DRY principles to ensure code quality and maintainability.